摘要
磁盘阵列是解决计算机Ｉ／Ｏ瓶颈问题的有效方法之一，通过对现有磁盘阵列结构的研究，提出了４种改善磁盘阵列性能的方法，即充分利用磁盘带宽，平衡多盘的负载，减少奇偶检验数据存取时间和磁盘阵列Ｃａｃｈｅ技术等。分析结果表明：在事务处理应用领域，利用Ｃａｃｈｅ来把小写转化为大写可大大改善外存的性能，平衡Ｃａｃｈｅ开销和阵列性能的提高是磁盘阵列设计的重点内容，在目前一般情况下，以一道数据大小来作为磁盘的基本存取大小是合适的选择，磁盘阵列负载平衡设计不仅是指正常模式下系统请求在多盘间负载平衡，而且在降级和重构模式下阵列内部的数据重构请求也应在多盘间平衡，利用备份磁盘作读写优化来减少奇偶检验数据存取时间是一种低Ｃａｃｈｅ开销。
Disk Array is an efficient method in solving the I/O bottleneck of the computer system. In this paper, the RAID architectures have been studied, and four methods to improve the performance of RAID, which are to use high throughput of single disk, to balance the I/O loads among disks, to reduce the response time of operating parity, and cache, are proposed. The results present: in business environment large data block access to the disk overcoming small write problem by cache can improved the performance of disk array. To balance the performance of disk array and the cost of cache is the important topic of ongoing research. For today′s typical disk parameter, it is the best to choose the basic access data block as a track data. Another result is that not only the I/O loads of computer system but also the reconstruction loads of disk array in degraded and rebuilt mode must balance. To decrease the access time the parity by the sparing disks is an inexpensive and efficient method to improve the performance of disk array.
出处
《计算机工程与设计》
CSCD
北大核心
1997年第3期20-27,共8页
Computer Engineering and Design
基金
国防科工委预先研究基金
关键词
磁盘
高速缓存
负载
磁存贮器
Magnetic disk Performance Array Cache Load
